---
title: Hypercyclic Toeplitz Operators
url: https://www.emergentmind.com/topics/hypercyclic-toeplitz-operators
type: topic
---

# Hypercyclic Toeplitz Operators

Hypercyclic Toeplitz operators are Toeplitz operators whose forward iterates admit a dense orbit in the underlying function space. On the Hardy space \(H^2(\mathbb D)\), and more generally on \(H^p(\mathbb D)\) for \(1<p<\infty\), their study lies at the intersection of linear dynamics, spectral theory, and function theory: hypercyclicity is controlled by the geometry of the symbol, the location of spectral components relative to the unit circle, the valence or winding of the symbol map, and the completeness of families of eigenvectors or reproducing kernels. The subject includes complete characterizations in several classical classes, near-sharp criteria for mixed analytic–antianalytic symbols, constructions of hypercyclic subspaces, weakly hypercyclic analogues, and extensions to de Branges–Rovnyak spaces and other operator models [2508.18874] [1506.06421] [1309.7627] [2502.03303].

## 1. Functional setting and dynamical notions

On \(H^2(\mathbb D)\), the Toeplitz operator with symbol \(\phi\in L^\infty(\mathbb T)\) is
\[
T_\phi f=P_+(\phi f),
\]
where \(P_+\) is the orthogonal projection \(L^2(\mathbb T)\to H^2\). The Hardy space itself is
\[
H^2(\mathbb D)=\left\{f(z)=\sum_{n\ge0}a_n z^n:\ (a_n)\in \ell^2\right\},
\]
the reproducing kernel at \(\lambda\in\mathbb D\) is
\[
k_\lambda(z)=\frac{1}{1-\overline{\lambda}z},
\]
and the Brown–Halmos theorem gives \(\|T_\phi\|=\|\phi\|_\infty\), while \(T_\phi^*=T_{\overline{\phi}}\) [2508.18874].

For a bounded linear operator \(T\) on a separable Banach space, hypercyclicity means that there exists \(x\) such that
\[
\{T^n x:n\ge0\}
\]
is dense. A hypercyclic subspace is an infinite-dimensional closed subspace every nonzero vector of which is hypercyclic. Weak hypercyclicity requires density only in the weak topology, and \(n\)-weak hypercyclicity requires that for every surjective continuous linear map \(S:X\to\mathbb C^n\), the projected orbit \(\{S(T^m x):m\in\mathbb N\}\) is dense in \(\mathbb C^n\) [1210.3191].

Several obstructions recur throughout the Toeplitz literature. If \(\|T\|\le1\), then \(T\) is not hypercyclic; if \(\|Tx\|\ge\|x\|\) for all \(x\), then \(T\) is not hypercyclic; if \(T\) is hypercyclic on a complex separable Banach space, then \(\sigma_p(T^*)=\varnothing\); and every connected component of \(\sigma(T)\) must meet \(\mathbb T\) [2508.18874]. These constraints are often converted into symbol conditions for concrete Toeplitz classes.

## 2. Classical characterizations on the Hardy space

The foundational Hardy-space result is the Godefroy–Shapiro characterization for anti-analytic symbols. If \(\phi\in H^\infty\), then
\[
T_{\overline{\phi}} \text{ is hypercyclic on } H^2
\iff
\phi \text{ is non-constant and } \phi(\mathbb D)\cap\mathbb T\neq\varnothing
\]
[2508.18874]. The mechanism is eigenvector-rich: for each \(\lambda\in\mathbb D\),
\[
T_{\overline{\phi}}k_\lambda=\overline{\phi(\lambda)}\,k_\lambda,
\]
so one obtains dense spans of eigenvectors with eigenvalues inside and outside the unit disk, and then applies the Godefroy–Shapiro criterion. This theorem recovers Rolewicz’s theorem for the backward shift as the case \(T_{\lambda\overline z}\).

A second classical family consists of tridiagonal Toeplitz operators
\[
T_F=aS^*+bI+cS,
\qquad
F(e^{i\theta})=ae^{-i\theta}+b+ce^{i\theta},
\]
with \(c\neq0\). Shkarin’s characterization states that \(T_F\) is hypercyclic on \(H^2\) if and only if \(|a|>|c|\) and the interior \(\mathcal E\) of the ellipse \(F(\mathbb T)\) intersects the unit circle:
\[
\mathcal E\cap\mathbb T\neq\varnothing
\]
[2508.18874]. Necessity of \(|a|>|c|\) is obtained from
\[
T_F^*T_F-T_FT_F^*=(|c|^2-|a|^2)(I-SS^*),
\]
since \(|c|\ge|a|\) makes \(T_F\) hyponormal, and a hyponormal operator on a Hilbert space cannot be hypercyclic. Sufficiency again relies on producing large families of eigenvectors and verifying their density.

The 2015 analysis of mixed symbols revisits this tridiagonal case and notes that Shkarin’s originally stated condition using
\[
\min_{z\in\mathbb T}|\Phi(z)|<1<\max_{z\in\mathbb T}|\Phi(z)|
\]
was incorrect, and that the correct condition is the one involving the complement of \(\Phi(\mathbb D)\) [1506.06421]. This correction is representative of the subject: geometric placement of the symbol image is decisive, but the correct geometric object is often subtler than a pointwise boundary modulus test.

## 3. Mixed analytic–antianalytic symbols and valence theory

A major extension of the classical theory studies Toeplitz operators with symbols
\[
\Phi(z)=p(\overline z)+\phi(z),
\]
where \(p\) is a polynomial and \(\phi\in H^\infty(\mathbb D)\) [1506.06421]. If \(\deg p=N\), the central geometric notion is \(N\)-valence: \(\Phi\) is \(N\)-valent in \(\mathbb D\setminus\{0\}\) if every equation \(\Phi(z)=w\) has at most \(N\) solutions in \(\mathbb D\), counted with multiplicity. The related set
\[
\Phi(\mathbb D,N)=\{w\in\mathbb C:\ \Phi(z)=w \text{ has exactly } N \text{ solutions in } \mathbb D\}
\]
encodes the values attained with maximal valence.

For \(\deg p=1\), the necessary conditions for hypercyclicity require univalence of \(\Phi\) in \(\mathbb D\setminus\{0\}\) together with boundary and spectrum conditions formulated in terms of \(\mathbb C\setminus\Phi(\mathbb D)\). For general \(N\), the necessary condition becomes \(N\)-valence in \(\mathbb D\setminus\{0\}\), and the relevant spectral set is \(\mathbb C\setminus\Phi(\mathbb D,N)\), not merely \(\mathbb C\setminus\Phi(\mathbb D)\) [1506.06421]. In the same paper, the sufficiency results assume \(\phi\in A(\mathbb D)\) and require either univalence up to the boundary in the degree-one case or exact maximal valence throughout the range in the general case.

The underlying spectral structure is unusually explicit. If \(\Phi\) is \(N\)-valent in \(\mathbb D\), then
\[
\sigma(T_\Phi)=\mathbb C\setminus \Phi(\mathbb D,N),
\qquad
\sigma_p(T_\Phi)\supset \mathbb C\setminus \Phi(\mathbb D),
\]
and for \(\lambda\in \mathbb C\setminus \Phi(\mathbb D)\) the eigenspace has dimension \(N\), with eigenvectors
\[
f_\lambda(z)=\frac{q(z)}{z^N\Phi(z)-\lambda z^N},
\]
where \(q\) is any polynomial of degree at most \(N-1\) [1506.06421]. Hypercyclicity is then reduced to completeness of eigenvector families. The key completeness theorem states that if \(h\in A(\mathbb D)\) is injective in \(\mathbb D\), then \(\{h^k\}_{k\ge0}\) is complete in \(H^2(\mathbb D)\); if \(h\in A(\mathbb D)\) is \(N\)-valent and each \(w\in h(\mathbb D)\) has exactly \(N\) preimages, then
\[
\{z^j h^k : k\ge0,\ j=0,1,\dots,N-1\}
\]
is complete in \(H^2(\mathbb D)\) [1506.06421]. Mergelyan’s theorem is the main approximation tool behind this density mechanism.

This framework introduced valence as a governing dynamical invariant for Toeplitz hypercyclicity. A plausible implication is that, in mixed-symbol problems, orbit structure is often encoded less by direct iterate estimates than by the covering behavior of the symbol map.

## 4. Rational antianalytic parts and smooth-symbol generalizations

The rational-symbol extension considers
\[
\Phi(z)=R(\overline z)+\phi(z),
\qquad
\phi\in H^\infty(\mathbb D),
\]
where \(R\) is rational with no poles in \(\mathbb D\) [2005.09557]. Writing \(N\) for the degree of the rational part, the paper proves that hypercyclicity forces \(\Phi\) to be \(N\)-valent in \(\mathbb D\). Its main reduction is to cyclicity for analytic multiplication operators: for \(\lambda\in\mathbb C\setminus \Phi(\mathbb D)\),
\[
h_\lambda(z)=\frac{1}{\Phi(z)-\lambda},
\]
and hypercyclicity of \(T_\Phi\) follows if the family \(\{1,z,\dots,z^{N-1}\}\) is cyclic for multiplication by suitable \(h_\lambda\) corresponding to spectral values on both sides of the unit circle [2005.09557]. This yields three sufficient conditions: the Maximal Valence Condition (MVC), the Increasing Argument Condition (IAC), and the Decreasing Valence Condition (DVC). The genuinely new feature is that IAC and DVC produce hypercyclic operators even when the symbol does not have constant valence, via deep cyclicity theorems of Solomyak.

A later model-theoretic generalization treats smooth symbols \(F\in C^{1+\varepsilon}(\mathbb T)\) on \(H^p\), \(1<p<\infty\), under the assumptions
\[
\varepsilon>\max(1/p,1/q),
\qquad
\frac1p+\frac1q=1,
\qquad
F'(\tau)\neq0 \text{ on } \mathbb T,
\]
together with a negative-winding hypothesis
\[
\w_F(\lambda)\le0
\qquad
(\lambda\in\mathbb C\setminus F(\mathbb T))
\]
[2502.03303]. Using Yakubovich’s model, for each \(\lambda\notin F(\mathbb T)\) one constructs eigenvectors
\[
h_{\lambda,j}(z)=z^j\frac{F_\lambda^+(0)}{F_\lambda^+(z)},
\qquad
0\le j<|\w_F(\lambda)|,
\]
and obtains a similarity
\[
T_F^*=U^{-1}M_\lambda U
\]
to multiplication by the independent variable on a vector-valued Smirnov space [2502.03303]. The resulting density theory leads to an orientation obstruction—if \(\w_F(\lambda_0)>0\) for some \(\lambda_0\notin F(\mathbb T)\), then \(T_F\) is not hypercyclic—and to exact criteria under geometric hypotheses: if every adjacent pair of spectral components satisfies either condition \((G)\) or the Jordan-curve alternative, then
\[
T_F \text{ is hypercyclic on }H^p
\iff
\Theta\cap\mathbb T\neq\varnothing
\text{ for every connected component }\Theta\text{ of }\overset{\circ}{\sigma(T_F)}
\]
[2502.03303].

| Symbol class | Structural condition | Dynamical conclusion |
|---|---|---|
| \(p(\overline z)+\phi(z)\) | univalence or exact \(N\)-valence | hypercyclicity via dense eigenvector spans |
| \(R(\overline z)+\phi(z)\) | MVC, IAC, or DVC | new hypercyclic classes, including varying valence |
| smooth \(F\in C^{1+\varepsilon}(\mathbb T)\) on \(H^p\) | negative winding and component criteria | necessary, sufficient, and exact hypercyclicity criteria |

The smooth-symbol theory recovers Shkarin’s ellipse case for all \(p>1\):
\[
T_F \text{ hypercyclic on }H^p
\iff
|a|>|c|
\ \text{and}\
\operatorname{int}(F(\mathbb T))\cap\mathbb T\neq\varnothing
\]
[2502.03303]. This situates earlier Hardy-space classifications inside a broader spectral-component framework.

## 5. Hypercyclic subspaces and weak hypercyclicity

Beyond existence of a dense orbit, one can ask whether a Toeplitz operator possesses a hypercyclic subspace. For the backward shift
\[
S(x_0,x_1,x_2,\dots)=(x_1,x_2,\dots)
\]
on \(\ell^2(\mathbb N_0)\cong H^2(\mathbb D)\), a 2013 paper studies operators \(\varphi(S)\), viewed there as coanalytic Toeplitz operators with analytic symbol \(\varphi\in A(\mathbb D)\) [1309.7627]. If
\[
\varphi(\mathbb T)\cap \mathbb T=\varnothing
\qquad\text{and}\qquad
\varphi(\mathbb D)\cap \mathbb T=\varnothing,
\]
then \(\varphi(S)\) has a hypercyclic subspace [1309.7627]. The proof has two parts. First, one uses the point spectrum
\[
\sigma_p(S)=\{\lambda\in\mathbb C: |\lambda|<1\}
\]
and the eigenvectors \((1,\lambda,\lambda^2,\dots)\), or equivalently the Cauchy kernels, to show that \(\varphi(S)\) is hereditarily hypercyclic via the Godefroy–Shapiro criterion. Second, one verifies
\[
\sigma_e(\varphi(S))\cap \overline{\mathbb D}\neq\varnothing
\]
using \(\sigma_e(S)=\mathbb T\), polynomial approximation in \(A(\mathbb D)\), and the Gonzalez–León-Saavedra–Montes-Rodríguez theorem, which converts hereditary hypercyclicity plus an essential-spectrum condition into a hypercyclic subspace [1309.7627]. The paper also emphasizes that the classical family
\[
\varphi(z)=\lambda z,
\qquad |\lambda|>1,
\]
is not covered, and that Montes-Rodríguez had shown earlier that \(\lambda S\) has no hypercyclic subspaces.

Weak hypercyclicity forms a parallel theory. A general criterion is given in terms of a dense \(T\)-invariant subspace carrying a stronger Hilbertian norm, an isometric extension with no nontrivial finite-dimensional invariant subspaces, and backward orbits converging to \(0\) [1210.3191]. Applied to coanalytic Toeplitz operators, it yields: if \(g\in H^\infty(\mathbb D)\) is non-constant,
\[
g(\mathbb D)\cap\mathbb D=\varnothing,
\]
and
\[
\{z\in\mathbb T: |g(z)|=1\}
\]
has positive Lebesgue measure, then \(T_g^*\) is weakly hypercyclic on \(H^2(\mathbb D)\) [1210.3191]. Conversely, if
\[
g(\mathbb D)\cap\mathbb D=\varnothing,
\qquad
|g|>1 \text{ a.e. on } \mathbb T,
\qquad
\log(|g|-1)\in L^1(\mathbb T),
\]
then \(T_g^*\) is not \(1\)-weakly hypercyclic, hence not weakly hypercyclic; moreover, for every nonzero \(f\in H^2(\mathbb D)\) and every \(k>0\),
\[
\lim_{n\to\infty} n^{-k}\,\|(T_g^*)^n f\|=+\infty
\]
[1210.3191]. The same paper proves that, on separable infinite-dimensional Banach spaces, weak hypercyclicity is equivalent to \(n\)-weak hypercyclicity for every \(n\in\mathbb N\).

## 6. Other function spaces and related operator models

The Hardy-space picture extends, with substantial modification, to de Branges–Rovnyak spaces \(H(b)\). Here \(b\) lies in the closed unit ball of \(H^\infty\), and \(H(b)\) is the reproducing kernel Hilbert space with kernel
\[
k_\lambda^b(z)=\frac{1-\overline{b(\lambda)}\,b(z)}{1-\overline{\lambda}z}.
\]
The fundamental dichotomy is between non-extreme \(b\), characterized by
\[
\log(1-|b|)\in L^1(\mathbb T),
\]
and extreme \(b\) [1812.07432]. For non-extreme \(b\), the paper states that \(T_{\overline\varphi}\) is hypercyclic on \(H(b)\) if and only if \(\varphi\) is non-constant and
\[
\varphi(\mathbb D)\cap\mathbb T=\varnothing
\]
[1812.07432]. It also proves that for the backward shift \(X_b=T_{\overline z}\), every scalar multiple \(\lambda X_b\) with \(|\lambda|>1\) is frequently hypercyclic, and that there is a dense \(G_\delta\) set of vectors common to all \(\lambda X_b\) for \(|\lambda|>\|X_b\|\) [1812.07432]. In the extreme case, by contrast, \(\lambda X_b\) is never hypercyclic for any \(\lambda\in\mathbb C\), and any hypercyclic \(T_{\overline\varphi}\) would have to satisfy
\[
\sigma_p(T_{\overline\varphi})=\varnothing
\]
[1812.07432].

A distinct but closely related model-space problem concerns truncated Toeplitz operators \(A_\psi\) on
\[
K_\theta=H^2\ominus \theta H^2,
\]
where \(\theta\) is inner [2112.08813]. The central open problem is explicit: do there exist hypercyclic truncated Toeplitz operators? For symbols
\[
\Phi(z)=a\overline z+b+cz
\]
and, more generally,
\[
\Phi(z)=\sum_{k=1}^{N} a_k z^{-k}+\sum_{\ell=0}^{M} c_\ell z^\ell,
\]
the paper computes point spectra and eigenfunctions in terms of the zeros of \(z^N(\Phi(z)-\lambda)\) [2112.08813]. In the three-term case, if \(|a|\neq|c|\) and an auxiliary function \(\Psi\) has no singular inner factor in the relevant Smirnov class on an annulus \(R_\beta\), then the set of eigenvectors of \(A_\Phi\) is complete in \(K_\theta\), and consequently \(A_\Phi\) is not hypercyclic [2112.08813]. Thus, unlike the classical Toeplitz setting on \(H^2\), the truncated theory presently supplies strong negative results and an unresolved existence question rather than positive classifications.

## 7. Conceptual themes, peripheral spectral links, and open problems

Several structural principles recur across the subject. The first is the unit-circle crossing principle: in the anti-analytic case it appears as
\[
\phi(\mathbb D)\cap\mathbb T\neq\varnothing,
\]
while in the smooth-symbol theory it becomes the requirement that every connected component of the interior of the spectrum intersect \(\mathbb T\) [2508.18874] [2502.03303]. The second is eigenvector propagation: reproducing kernels, explicit resolvent-type eigenfunctions, and model-theoretic eigenvector bases are used to manufacture dense linear spans. The third is symbol geometry: univalence, \(N\)-valence, winding numbers, and adjacency of spectral components are not auxiliary hypotheses but the main dynamical invariants.

Current open problems reflect exactly these themes. In the smooth-symbol framework, the literature asks whether every hypercyclic Toeplitz operator must satisfy the Godefroy–Shapiro criterion, whether the inclusion
\[
H_{\Omega'}(T_F)\subseteq H_\Omega(T_F)
\]
always holds when \(|\w_F(\Omega)|<|\w_F(\Omega')|\), whether hypercyclicity depends only on the geometric curve \(F(\mathbb T)\) rather than its parametrization, and whether hypercyclicity is independent of the Hardy exponent \(p>1\) [2502.03303]. In the truncated setting, the overarching open problem remains whether any hypercyclic truncated Toeplitz operators exist at all [2112.08813].

There is also an indirect spectral connection with Weyl-type theorems. A 2024 paper does not study hypercyclic Toeplitz operators as a standalone topic, but shows that for a Toeplitz operator of the form \(T_{z+q}\) on the Bergman space, if it happens to be hypercyclic or supercyclic, then it satisfies property \((UW)_E\); more generally, for hypercyclic operators \(A\), property \((UW)_E\) is equivalent to \(E(A)=\varnothing\) [2410.08838]. This suggests that hypercyclic Toeplitz theory interfaces not only with linear dynamics and function theory, but also with finer Weyl-type spectral identities, although in that paper the connection is explicitly indirect rather than classificatory.

Source: https://www.emergentmind.com/topics/hypercyclic-toeplitz-operators